HR-Glossar
HR-Analytics (people analytics)
What HR analytics can deliver, where German data protection and co-determination draw the line and why individual predictions are delicate.
1. What is HR-Analytics (people analytics)?
HR analytics means the systematic analysis of personnel data to identify relationships and support decisions – on turnover, time to fill, absence, training or pay structure, for instance.
The difference from HR controlling lies in the ambition. Controlling reports measures; HR analytics looks for relationships and sets out to explain why something is as it is. It is precisely that ambition which creates the legal questions.
Two limits therefore have to be settled in advance, not afterwards. First data protection: processing employee data is tied to necessity, and consent mostly does not hold in a relationship of dependence. Second co-determination: an analysis system is regularly a technical device capable of monitoring conduct or performance – and therefore subject to co-determination under Section 87 BetrVG, regardless of intention.
2. Origin and development
Personnel data have always accumulated but were largely administered rather than analysed. As payroll, time recording and applicant management were digitised, analysable stocks arose – and with them the question of what could be learnt from them.
Expectations were high at first: if relationships could be shown, it should also be possible to predict who will resign, who will succeed, who will be absent. That expectation has been met only in part, for two reasons.
The first is methodological. Personnel data are rarely clean, case numbers in most businesses are small, and the relationships observed are correlations. Drawing conclusions about individuals from them overstretches the data – with statements that are often wrong in the individual case.
The second is legal. A prediction about a person is a processing of personal data with considerable effect. It has to be measured against necessity, and where it prepares automated decisions further questions arise.
What actually holds is more modest and more useful: analysis at the aggregate level. Where turnover clusters, where hiring fails, where pay differences exist – those are insights about the organisation, not about people.
3. Core principles and how it works
Necessity limits the processing
Not everything that could be analysed may be analysed. The measure is necessity for the employment relationship.
Consent rarely holds here
In a relationship of dependence, freedom is doubtful, and a withdrawal would make the system unusable.
Co-determination because of the capability to monitor
An analysis system is regularly subject to co-determination under Section 87 BetrVG – the capability to monitor suffices, no intention is required.
Aggregate rather than individual
Analysis across groups and periods is sound in law and in method. Statements about individuals mostly are not.
Correlation is not cause
A relationship shows where to look more closely – not what to do. The step to a cause needs other methods, usually conversations.
Data quality limits everything else
Inconsistently maintained master data produce analyses that look precise and are wrong.
Transparency towards those concerned
Which data are analysed for what purpose belongs made known. Covert analysis damages more than any insight is worth.
4. Who is HR-Analytics (people analytics) relevant for?
- HR functions with a steering role – for them it is the route from instinct to evidence. - Management – underpinning personnel decisions with figures changes the discussion. - Works councils – analysis systems are subject to co-determination, and how they are arranged is negotiable. - Data protection leads – purpose limitation, necessity and erasure have to be settled. - IT leads – data quality and interfaces determine how much the results are worth.
5. How it differs from related terms
- HR analytics and HR controlling – controlling reports measures, analytics looks for relationships. The boundary is fluid, the legal effort is not. - Aggregate and personal analysis – the decisive distinction. Aggregates are about the organisation, individual analyses about people. - HR analytics and performance appraisal – a data-supported assessment of people is something other than analysing organisational patterns; mixing them is delicate. - HR analytics and algorithmic pre-selection – in a hiring process AGG questions are added: a system trained on historical hiring reproduces its patterns. - Measure and target – a measure that becomes a target stops being a good measure. That applies particularly here, because the metrics are easy to influence.
6. Variants and adaptations
Applications, ordered by how well they hold:
- Turnover analysis by area and joining cohort – aggregated, telling, unproblematic in law. The best place to start. - Analysis of hiring processes – where procedures break off, how long they take. - Pay structure analysis – makes indirect discrimination visible; sensible in law and in part required. - Absence analysis – permissible in aggregate; at individual level health data are involved and the requirements are considerably higher. - The effect of training – methodologically demanding, because comparison groups are missing. - Individual resignation prediction – methodologically delicate, legally challengeable and risky in its effect on trust.
7. Advantages and challenges
Advantages
- Replaces assumptions with evidence, particularly on turnover and hiring problems
- Makes indirect discrimination visible in pay and promotion structures
- Aggregated analysis is unproblematic in law and often already telling
- A works agreement creates legal certainty for the whole system
- Forces better data quality, which benefits every process
Challenges
- Personal predictions are methodologically weak and legally challengeable
- Small case numbers produce spurious relationships that look convincing
- Poor master data quality produces wrong results that look precise
- Co-determination and data protection are regularly considered only after the system is chosen
- Covert analysis damages trust more lastingly than any insight is worth
- Correlations are read as causes and lead to measures at the wrong point
8. Best practices for implementation
Start with aggregated analysis
Turnover by area, time to fill, pay structure. That is unproblematic in law, sound in method and answers most of the questions you actually have.
Settle co-determination before choosing the system
An analysis system is regularly subject to co-determination under Section 87 BetrVG. Noticing that only before the rollout costs months.
Use a works agreement as the basis
It legitimises the processing and satisfies co-determination in one step – and it is more robust than any consent.
Test correlations with conversations
A striking relationship shows where to look. What lies behind it is usually learnt in conversation, not in the data.
Make it transparent what is analysed
Employees should know what analyses exist and what for. It costs little and prevents the suspicion that otherwise arises.
9. Tips for employers and employees
For employers
- **Start aggregated** – that answers most questions with no legal risk
- **Section 87 BetrVG before choosing the system** – capability to monitor suffices, intention is not needed
- **A works agreement rather than consent** – more robust and it satisfies both
- **Data quality first** – poor data produce wrong results that look precise
For employees
- **You can request access** – Article 15 GDPR, including analyses relating to you
- **The works council is involved** – for systems capable of recording conduct or performance
- **Consent is revocable** – and a withdrawal must not harm you
- **Ask about the purpose** – analyses with no discernible purpose deserve questioning
10. Conclusion
HR analytics differs from HR controlling in its ambition: not reporting measures but explaining relationships. It is from that ambition that the limits arise, and they belong at the start – not in an annex.
In law there are two. Processing employee data is tied to necessity, and consent mostly does not hold in a relationship of dependence. And an analysis system is regularly subject to co-determination under Section 87 BetrVG, because the capability to monitor suffices – intention does not matter. A works agreement solves both in one step. For a group rolling out a central people analytics platform, that is the step that cannot be skipped, however the tool is meant.
Methodologically the decisive line runs between aggregated and personal analysis. Aggregates across areas and periods hold up and answer most of the questions one actually has: where turnover clusters, where hiring fails, where pay differences exist. Predictions about individuals, at the case numbers usually available, are methodologically weak, legally challengeable – and in their effect on trust more expensive than any insight they deliver.
Sources
- Section 26 BDSG – Data processing for the purposes of the employment relationship (official English version, as at 23 June 2021; the German text governs) (opens in a new tab)
- Section 87 BetrVG – Co-determination rights in social matters (official English version) (opens in a new tab)
- Regulation (EU) 2016/679 (GDPR) (English language version) (opens in a new tab)
- Federal Office for Information Security (English pages) (opens in a new tab)
Related terms
Our promise
Software supports. People take responsibility.
Let us talk about your payroll – no strings attached, specific, and with a dedicated contact from day one.